摘要
提出了一种在时间与面积约束下,运用遗传算法同时进行操作调度和资源分配的高层次多电压功耗优化方法.对于时间和面积约束所导致的无效染色体,通过将约束优化问题转换成两个目标函数的极值问题,一个为原问题的目标函数,另一个为违反约束条件的程度函数,避免了约束条件对问题求解的影响.对于数据依赖所导致的无效染色体,采用基于数据依赖的单点杂交算子来解决.实验结果表明,该算法比不考虑无效染色体处理机制的简单遗传算法的多电压功耗优化方法的功耗优化能力提高10%,收敛速度提高15%.
This paper proposes a time-and-area-constrained high-level power optimization method for a multiple supply voltage, in which operation scheduling and functional unit allocating are simultaneously done through the genetic algorithm. Invalid chromosomes caused by time and area constraint are avoided by treating constraint optimization as a two-objective optimization function： one objective is the original objective function and the other is the degree violating the constraint conditions, thereby avoiding the influence of constraint conditions on the problem solving. The problem of invalid chromosomes caused by data dependence is solved through one point crossover operator based on data dependence. Experimental results show that this algorithm has improved power optimization by 10% and the convergence rate by 15~, compared with those simple power optimization methods for a multiple supply voltage of the genetic algorithm which does not consider invalid chromosomes handling.
出处
《西安电子科技大学学报》
EI
CAS
CSCD
北大核心
2009年第5期933-939,共7页
Journal of Xidian University
基金
国家自然科学基金资助(60273081)
关键词
低功耗
高层次综合
多电压
遗传算法
low power
high-level synthesis
multiple voltages
genetic algorithm
